Biomedical Image Geometric Morphometric Study

The transformation of any organism in morphological structure reflected the changes in their function. The geometric morphous of organisms could be measured through biomedical images to gain the feature information and activity rules quantitatively. It was helpful to assist medical diagnosis and treatment. With the development of medical imaging and computer technologies, three-dimensional models were reconstructed by tomographic imaging technology which contained the outer side and inside stereo structure. The images included whole morphous information, which provided a good platform for geometric morphological measurement. However, this modeling method needed expensive equipments and the imaging method was not fitted for the studies in some biomedical area. According to the research purposes and objects, biomedical images were still generated by traditional methods in many conditions. The incomplete morphological information increased the difficulty of accurate measurement. The estimation and fitting could increase the accuracy and efficiency of the measurements.In this work we did the research about the geometric morphological measurements of microscopic images and anatomical images, including the two-dimensional slice images of rats' testicle tissues and three-dimensional surface models of human breasts. The lost morphological information were estimated and fitted with the improved method in the research according to the existed problems in the present methods. The research increased the accuracy and efficiency of the methods.Two-dimensional slice images of rats' testicle tissues only contained the information of intersecting surfaces. The three-dimensional parameters were measured by using stereological method. The basic theory of stereological method was estimating the two and three dimensional parameters such as area and volume by covering the geometric probes on the slice images. The stereological measurement program, SMP, was developed to provide the complete functions of stereological measurement. Partial results were significant different with actual values and the efficiency of manual probes counting was low with traditional stereogogical methods. To overcome these problems, an automatic measurement method was proposed based on image segmentation. The stereological measurement could be much more efficiency and accurate by the program of automatic porbes counting. The biomedical images with simple color distributions could be segmented by threshold methods. However, the slice images of rats' testicle tissues were noisy and complicated in color and morphology. The segementation results of traditional methods usually unsatisfactory. In this paper we proposed a novel image segmentation method based on multiple image processing and segmentation methods. The noises were removed by using the high order statistical method. Then the method with the combination of multiple thresholds and morphological models segmentation achieved the automatic segmentation of seminiferous tubule in the images.The lost imformation of three-dimensional surface models of the breast were internal morphological information. Thus the definition of breast base was the key point of the volume measurement. The usual method with defining a base plane was bad in repeatability and did not consider the effects of body posture and the chest contour. A novel method was proposed to fit the surface of chest wall by the four border curves of the breasts. The volume results witch calculated by this method was more close to the actual values. On the other side, there was no standard index to evaluate the symmetry of breasts. A new evaluation index which was the ratio between the volumes intersection and union of the two side breasts was proposed. It was a better parameter to measure the symmetry of breasts.At last, an online three-dimensional measurement system was builted based on the patients' information database from the plastic surgery. The users could observe the breasts' three-dimensional models and measure the linear parameters, volumes and symmetry index through web browsers directly.The major innovations in this paper were summarized below.1. A novel image segmentation method was proposed. The hide information was extracted by the high order statistical method. The seminiferous tubule profile in rats' testicle slice images were automatically segmented according to the multiple thresholds method and morphological models built by priori knowledge.2. A measurement program was developed according to stereological theory to measure the geometric morphous through two-dimensional images. The geometric probes could customize and counted automatically based on the segmentation result, which improved the efficiency and accuracy of the measurement.3. A novel method was proposed to fit the chest wall surface with four border curves of breast. The result was more close to the actural volumes of the breasts.4. A new breast symmetry index was proposed. It could be more accurate to measure the symmetry of two side breasts than the previous RMS value.5. An online system was developed to measure the three-dimension models of the breasts. Users could visit the system directly through web browsers to measure the morphological parameters of the breasts.